Semiconductor optoelectronic devices are efficient energy transformers which can convert electrical energy into optical energy. The reciprocal process is also possible wherein optical energy is converted directly into electrical energy. Optically emissive devices that generate light from an electrical source are used as solid-state semiconductor light sources such as incoherent light emitting diodes (LEDs) and coherent light emitting laser diodes. Semiconductor light sources take advantage of the interaction of optical energy with the semiconductor's crystal structure which has a specific electronic energy configuration known as the electronic band structure.
The industrial and commercial potential for LEDs and semiconductor lasers continues to expand. Although LEDs have been available since the 1960s, the efficiency of LEDs has continuously improved since that time, and solid-state lighting applications using LEDs are now considered a viable alternative to incandescent and high intensity discharge light sources. Modern LEDs produce light across the visible and near infrared electromagnetic spectrum, and researchers have developed semiconductor materials for LEDs that produce deep ultraviolet (DUV) light typically operating in the optical wavelength range of 190 nm to 290 nm. DUV-LEDs producing incoherent light have significant commercial potential for surface and air sterilization or sanitation, extremely high-density optical storage having high speed read & write ability that is competitive to nanoscale imprint technologies, fine geometry lithography, and information processing systems generally. In particular, optical communication links using the deep ultraviolet band offers unique short range device connectivity. Various additional applications that require efficient and compact sources for ultraviolet light are enabled by DUV-LEDs such as those involving the interrogation of biological systems.
Semiconductor light sources generate light using semiconductor junctions comprising at least a p-type semiconductor region and an n-type semiconductor region. The p-type semiconductor region is designed to be a source of holes, whereas the n-type region is a source of electrons. Under the appropriate external electrical bias, electron and holes are injected from their respective sources towards an electron-hole-recombination region (EHR)—which can be described with reference to FIG. 1a. 
For the case of a p-n junction diode, the electron-hole-recombination region is substantially the depletion region between the p-type and n-type semiconductors. Alternatively, an intentional intrinsic region or not-intentionally doped (NID) region can be inserted between the p-type and n-type regions to form a p-i-n diode. A p-i-n diode is designed such that a majority of the electron-hole-recombination occurs spatially within the intrinsic or NID region.
FIG. 1a illustrates a semiconductor junction 100 comprising distinct spatial regions of semiconductor material types and the associated energy band diagram 101 along a spatial direction comprising the ordinate of the diagram. The abscissa of energy band diagram 101 correlates spatially with cross section 102 of semiconductor junction 100. In other words, the left side of energy band diagram 101 corresponds to the p-type portion 103 of semiconductor junction 100, and the right side of energy band diagram 101 corresponds to the n-type portion 104 of semiconductor junction 100. The p-type portion 103 of the semiconductor junction comprises mobile positive charge carriers in the form of holes 105. The n-type portion 104 of the semiconductor junction comprises mobile negative charge carriers in the form of electrons 106. The electrons and holes can diffuse and drift toward the depletion region or intrinsic region 107. For the case of p-n and p-i-n junctions, the regions 103, 107 and 104 can comprise a uniform bandgap composition and can collectively be referred to as a homojunction. If anyone or all of the regions 103, 107 or 104 comprise dissimilar semiconductor compositions of differing bandgap, then a so called heterojunction device is formed. For the case of a p-n device, there will be an equilibrium reaction wherein electrons and holes diffuse across the interface defining a p-type region and an n-type region forming a built-in potential. This built-in potential establishes a depletion region that is neither n-type nor p-type and is ideal for functioning as an EHR. In p-i-n junctions, the region 107 is additionally and intentionally engineered with a predetermined dimension which is often larger than the depletion region formed by an abrupt p-n junction. This is particularly useful for controlling the spatial extent of EHR for the purpose of photon generation and extraction of said photon externally from the device.
The mechanism by which semiconductor junction 100 generates light comprising photons having distinct energy spectrum (or equivalent wavelength spectrum) is by the simultaneous particle recombination of spatially coincident electrons in the lowest energy portion of the conduction band with holes in the highest portion of the valence band. That is, the process of generating a photon of energy Ep is determined by the energy difference between an electron in the conduction band and hole in the valence band where Ep=Ec(k=0, z)−Ev(k=0, z)˜EG. The respective electron and hole energies are in general taken at the same semiconductor crystal momentum vector k of the host semiconductor band structure (having a distinct energy-momentum dispersion defining said band structure) in order to conserve energy and momentum during the conversion process. This is due to the fact that a photon has zero momentum. High efficiency optical generation via electron-hole recombination process occurs for so-called direct bandgap semiconductors wherein the energy-momentum dispersion has the lowest lying energy of conduction band occurring at k=0 and the highest lying valence band also occurring at k=0. In general, all optical properties of interest for light emission occur in the vicinity of the so called zone center of the band structure, which is centered at k=0. Therefore, the bandgap of the host semiconductor represented by EG≡EG(k=0) is typically defined as the energy difference between said lowest energy conduction and highest valence band energies at k=0. This allows representations of the spatial band structure of p-n and p-i-n devices to be abbreviated by their k=0 representation, as described with reference to FIG. 1a. Note, in general Ep≦EG due to the coulomb attraction of the electron and hole forming an intermediary particle called an exciton having binding energy EB.
The energy band diagram 101 of the layered p-i-n diode 100 of FIG. 1a is therefore representative of a homogeneous bandgap semiconductor spatially modified along a direction comprising the ordinate of the diagram, so as to exhibit a distinct p-type 103, an optional NID or depletion region 107 and an n-type region 104. The so called conduction band edge Ec(k=0, z) 108 represents the k=0 energy position of the lowest energy conduction band of the host semiconductor, whereas the valence band edge Ev(k=0, z) 109 represents the highest valence band energy position relative to the conduction band edge 108.
The p-type 103 and n-type 104 regions can be electrically contacted by ideal low resistance ohmic contacts forming a two electrical terminal device. Under appropriate externally applied electrical bias applied to the contacts, the junction of FIG. 1a can be operated to function in forward bias mode wherein mobile electrons 106 in region 104 and holes 105 in region 103 are injected toward region 107. For the case of a p-n homojunction, region 107 is the well known depletion region of width WD setup by the abrupt junction formed between the n-type 104 and p-type region 103. The built-in electric field along the device direction, comprising the ordinate of the diagram, is determined by the difference in acceptor and donor doping concentrations in regions 103 and 104, respectively, and the applied electrical bias.
For the case of a p-i-n diode, the intrinsic region has a width that is intentionally engineered to have width Wi and is generally not intentionally doped with impurity species or is advantageously compensated or chemically modified so that the Fermi energy EF lies wholly within the bandgap of the host semiconductor. Preferably, the intrinsic region should be devoid of non-radiative traps or loss paths provided by structural defects and have EF positioned approximately to half of EG.
Once the light has been generated in the semiconductor junction it will need to be coupled out from the interior of the device and emitted externally for utility as a light source. In this regard, the design of semiconductor laser diodes and LEDs differ significantly. Both LEDs and laser diodes can be designed to be of a vertically emitting type (i.e., light is generated within the device and emitted substantially perpendicular to the plane of the layers) or a waveguide type (i.e., light generated and guided within the device and emitted substantially parallel to the plane of the layers). In both of these optical design configurations, the optical polarization of the generated light is of critical importance.
In general, lasers require the additional feature of parallel reflective layers bounding the optical generation region in order to recycle photons back into the EHR region for the express purpose of producing stimulated emission and thus coherent light. In contrast, LEDs are generally classed as producing photons due to spontaneous emission process and thus produce incoherent light output. LEDs can produce light that is optically polarized or non-polarized, whereas laser diodes produce polarized emitted light. The polarization state of a given device is determined by both the semiconductor bandstructure and the optical structure configuration, namely, vertical emitter or waveguide type. The choice of any one particular optical configuration using specific semiconductor materials will be determined by the optical properties of the light required for a given application and the cost of producing the device. In general, vertically emissive devices are of lower cost than waveguide type devices. However, achieving high optical extraction efficiency (which is the ratio of the amount of light generated within the device to the amount of light extracted externally from the device), is often the prime goal. This has proven to be a limiting attribute for deep ultraviolet LEDs wherein the extraction efficiency is fundamentally limited by the semiconductor band structure and the high optical refractive index of the semiconductor material itself.
The effect of the optical polarization of light internally generated by a semiconductor junction on the performance of the device can be described with reference to FIG. 1b. This figure displays a semiconductor structure 110 comprising an active layer 111 and a substrate 112. Substrate 112 is generally a single crystal material if used for subsequent epitaxial deposition of active layers 111. Alternatively, substrate 112 may be non-crystalline or amorphous if used as a mechanical support or as an optical coupler for otherwise mechanically transferred crystalline active layers 111.
Substrate 112 may further be either selected from an optically opaque or transparent material if utilized as a crystalline template for seeding epitaxial layers 111. If the substrate is transparent to the design emission wavelength, then light can be coupled externally from the device layers 111. If the crystalline substrate 112 is opaque to the design wavelength, then optical energy can be coupled out from the device laterally as a waveguide or from the topmost surface. Yet a further method of coupling energy out from active layers 111 using an opaque substrate is to remove a portion of the substrate material beneath the active region of a device so as to enable emission through an aperture in the substrate.
For application of homojunction and heterojunction LEDs to deep ultraviolet wavelengths in the range of 200 to 260 nm, the wide band gap semiconductor materials having a composition of aluminium-gallium-indium-nitride (with chemical formula AlxGayIn1-x-yN, where 0≦x≦1, 0≦y≦1) and magnesium-zinc-oxide (MgxZn1-xO where 0≦x≦1, 0≦y≦1) are of particular interest. Other material compositions are also possible. However, the technologically mature AlGaInN and MgZnO materials have been shown to be able to form stable and continuous compositions of ternary and quaternary alloys by varying the mole fractions of the incorporated metals within the nitride or oxide crystalline structure.
The group-III Nitride material is presently the most mature wide band gap semiconductor material and is widely used in near ultraviolet and visible LEDs in the wavelength range of 250 to 600 nm. Unfortunately, deep ultraviolet LEDs based on AlGaN, AlInGaN or AlInN operating with bandgap energies in the range from about 200 nm to approximately 260 nm suffers a deleterious and fundamental limitation of the electronic band structure that is inefficient for vertical type emitters.
Furthermore, heterojunction LED devices using two or more dissimilar composition of group-III Nitrides formed as a wurtzite crystal structure suffer from the disadvantageous creation of internal spontaneous and piezoelectric electric fields due to extremely large induced polarization charges at each heterojunction interface. That is, at each interface (for example the interface between AlxGa1-xN/AlyGa1-yN where x≠y) these polarization charges generate built-in electric polarization fields which tend to prevent the efficient spatial localization of electrons and holes, and thus exhibit poor electrical-to-optical generation efficiency.
The internal polarization charges at a group-III Nitride heterojunction interface can be reduced and potentially eliminated if grown epitaxially as substantially single crystal films on semi-polar or non-polar surfaces, and have been identified in the prior art. Unfortunately, epitaxial deposition of thin films of wurtzite group-III Nitrides on non-polar crystal surfaces typically result in poor crystal quality films when compared to a deposition on the so called polar surface c-plane having hexagonal unit cell crystal symmetry. Thus the highest structural quality material for group-III Nitrides (and indeed also for MgZnO) is obtained by depositing on the polar c-plane and in turn results in the highest polar type of crystal with largest internal polarization charge for a given heterojunction interface. Thin film single crystal group-III Nitrides deposited along a growth direction which is substantially perpendicular to the c-plane therefore results in the two distinct crystal structure film types labelled as metal-polar and nitrogen-polar. Achieving a group-III Nitride structure that exhibits pure polarity-type, namely metal-polar and nitrogen polar is in general a desirable goal for device operation. Introduction of structural defects increases the probability of creating mixed polarity domains within the films during epitaxy.
There is yet a further problem with prior art group-III Nitride bulk substrate growth and epitaxial deposition methods. Native single crystal group-III Nitride binary substrates of GaN and AlN are available with low defect density but have relatively high cost of manufacture when compared to high quality single crystal sapphire and silicon. Epitaxial growth of electronic and optical devices using single crystal low defect density ternary AlGaN and AlInAl or quaternary AlInGaN films are also limited by the in-plane lattice constant mismatch Δa of the film with respect to the crystalline substrate. Therefore, Δa is a function of group-III Nitride epitaxial material composition relative to the underlying substrate lattice constant.
In general, epitaxial layer deposition degrades in structural quality for a given group-III Nitride composition when the individual layer thickness exceeds the so called critical layer thickness (CLT) when deposited on a dissimilar substrate. The critical layer thickness is defined as the maximum epitaxial layer thickness that can accommodate the in-plane lattice distortion of the crystal unit cell elastically without creating misfit dislocations. The CLT correlates strongly with the lattice mismatch. This places severe limitations on the material combinations of epitaxially layered heterojunction devices if structural defects are required to be low.
In particular, for the case of LEDs and laser diodes, these structural crystal defects arising from lattice constant mismatch of dissimilar group-III Nitride films present alternative electron and hole recombination pathways that are non-radiative in nature and thus represent a major production loss for the desired emission energy engineered by the bandgap or heterojunction device.
Therefore, the lattice constant mismatch problem of heterojunctions and multilayered stacks of dissimilar composition films (each having an associated distinct band structure and crystal lattice constant) limits the ultimate internal quantum efficiency and design range of emission wavelengths for optically emissive devices. The management of internal elastic strain of each of the dissimilar layers comprising a heterojunction device or multilayered stack comprising a plurality of dissimilar composition films is important for controlling the final crystal structure quality of the device. In practice, the previously discussed lattice mismatch issues place severe limitation on the choice compositions selected for constructing a given device and thus the design space of possible bandgap engineered devices.
Yet a further major problem in the group-III Nitrides is the deposition process complexity for creating multiple layers of dissimilar compositions with a wide range of alloy compositions within a given deposition process, albeit by using either chemical vapor deposition (CVD), metal organic CVD (MOCVD), molecular beam epitaxy (MBE), vapor phase epitaxy (VPE), sputter deposition, ion beam deposition, and other methods. That is, in general large variation in the composition of the group-III Nitrides alloys typically require distinct deposition process specification to achieve high structural quality films. For example, Al-rich films of a given AlxGa1-xN alloy (x>0.5) require different process conditions compared to Ga-rich films (x<0.5) such as different constituent metal-to-active nitrogen ratios during deposition and different growth temperatures. Therefore, high quality multilayered thin film epitaxial structures comprising a plurality of widely varying AlGaN film compositions require a complex growth process and thus have an increased cost and potentially lower final structure production yield due to normal deposition process variability.